Pennsylvania data centers face scrutiny over water and power usage

Pennsylvania is emerging as a prime target for artificial intelligence data centers, but a new tension is building over how these facilities will consume local electricity and water. The state's proximity to major East Coast cities and its existing energy infrastructure make it an attractive location for companies building out AI capabilities. However, the rapid expansion of these centers is straining the very resources that local communities rely on for daily life.

Brian R. King, an associate professor of computer science at Bucknell University, explains that these data centers house powerful computers designed to train and operate AI models. The hardware relies on graphics processing units, or GPUs, which demand substantial amounts of electricity. King notes that while this is often viewed as a technology shift, it is actually disrupting every field and domain. He argues that the energy requirements are outpacing the pace at which local grids can easily adapt.

The pressure is not limited to the power grid. Dr. Benjamin Hayes, a program director of Watershed Sciences and Engineering, points out that many large data centers use water to cool their equipment. This creates a dual burden on local infrastructure. The footprint of these projects includes water used for cooling, water used to generate the necessary electricity, and the physical conversion of farmland into industrial zones.

Hayes warns that the costs of this expansion will not stay contained within the companies building the facilities. Neighbors and taxpayers may face higher electricity bills and reduced availability of public water. He argues that local officials are currently evaluating proposals in isolation. Instead of looking at the big picture, they are approving projects one by one without accounting for how multiple data centers might impact the same watershed or power supply simultaneously.

Hayes advocates for a statewide or regional analysis that accounts for electricity and water demand, drought conditions, land use, and community needs. He notes that there is currently no central clearinghouse to review how these projects interact with each other. This lack of coordination means that the cumulative impact on local resources is often overlooked until the strain becomes visible.

King suggests that developing smaller AI models designed for specific tasks could help alleviate the pressure. Instead of relying exclusively on increasingly large models that require extensive computing resources, the industry could shift toward more efficient systems. Both researchers agree that energy use, water availability, and local infrastructure must remain at the center of the conversation as Pennsylvania continues to weigh the opportunities and challenges of data center development.